April 04, 2005

Sheep or Gulls? (This Week at the Complex Systems Colloquium)

Attention conservation notice: This is an attempt to increase the attendance at the complex systems colloquia by blogging about them in advance. Of minimal relevance if you're not in Ann Arbor or don't care about complex systems, modeling the spread of ideas and practices through social networks, herding and information cascades, fads and fashions, relating micro-actions to macro-behavior, or how to successfully launch one of those silly blog memes.

For this week's colloquium, we are very happy to have as our speaker Michelle Girvan from SFI. Michelle has done extremely impressive and well-known work on complex networks, perhaps most notably her algorithms for discovering community structure in networks. This week, however, she is going to answer for us a question asked by generations of parents, "If everybody else jumped off a bridge, would you jump too?", as well as its less time-honored companion, "What if they all told you it was 'gravitational therapy' and really good for your calves?"

"Persuasion, Imitation, and the Spread of Ideas in Social Networks"
Abstract: The spread of ideas and behaviors through populations depends crucially on the structure of the social network and the local mechanism of transmission between individuals. In this talk, I will explore how information cascades differ when transmission is based on persuasion versus imitation. In addressing the implications of the different modes of transmission, I take a network-based approached that focuses on the following types of questions: if a fad spreads from individual to individual through imitation, which nodes in the network are the best initial targets to maximize the size of the cascade? If the transmission mechanism is instead persuasive, how do we need to modify our target set? I will discuss two types of imitation — infectious imitation in which each link is given equal probability for transmission, and herd imitation in which the probability that an individual becomes an adopter depends on the fraction of his or her neighborhood that has already adopted the idea or behavior. I will also discuss a simple model of persuasion that is built around the assumption that all individuals have equal time to devote toward persuading their neighborhood and they attempt to divide the time equally among their associates. We see that the best initial target set changes substantially when the mode of transmission is altered. In addition, structural features of the social network heavily influence the advantage of an optimized versus random approach to choosing the initial target set.

Thursday, 7 April, at 4 pm in 335 West Hall, Central Campus.

UPDATE: Sadly, Dr. Girvan has had to cancel her visit to Ann Arbor. The talk is off, at least for this semester.

Complexity; Networks

Posted at April 04, 2005 08:45 | permanent link

Three-Toed Sloth